2025-08-07
GPT-5 (high)
by OpenAI
Max Tokens
128000
Competition performance
| Competition | Accuracy | Rank | Cost | Output Tokens |
|---|---|---|---|---|
|
IMProofBench - Proofs
🕵️ Research Math
|
50.23% ± 14.00% | 1/5 | N/A | N/A |
|
IMProofBench - Final Answers
🕵️ Research Math
|
52.93% ± 15.10% | 5/11 | N/A | N/A |
|
Apex
🏔️ Apex
|
1.04% ± 1.44% | 12/22 | $5.54 | 46122 |
|
Overall
👁️ Visual Mathematics
|
78.75% ± 2.97% | 4/13 | $2.04 | 7243 |
|
Kangaroo 2025 1-2
👁️ Visual Mathematics
|
68.75% ± 9.27% | 4/13 | $1.52 | 6222 |
|
Kangaroo 2025 3-4
👁️ Visual Mathematics
|
60.42% ± 9.78% | 8/13 | $2.39 | 9838 |
|
Kangaroo 2025 5-6
👁️ Visual Mathematics
|
65.00% ± 8.53% | 8/13 | $2.41 | 7952 |
|
Kangaroo 2025 7-8
👁️ Visual Mathematics
|
90.83% ± 5.16% | 2/13 | $1.96 | 6449 |
|
Kangaroo 2025 9-10
👁️ Visual Mathematics
|
92.50% ± 4.71% | 7/13 | $1.72 | 5632 |
|
Kangaroo 2025 11-12
👁️ Visual Mathematics
|
95.00% ± 3.90% | 4/13 | $2.24 | 7363 |
|
Overall
🔢 Final-Answer Competitions
|
91.02% ± 1.96% | 8/17 | $5.04 | 14156 |
|
AIME 2025
🔢 Final-Answer Competitions
|
95.00% ± 3.90% | 4/54 | $4.08 | 13475 |
|
HMMT Feb 2025
🔢 Final-Answer Competitions
|
88.33% ± 5.74% | 15/54 | $5.00 | 16380 |
|
BRUMO 2025
🔢 Final-Answer Competitions
|
91.67% ± 4.95% | 16/40 | $3.28 | 10760 |
|
SMT 2025
🔢 Final-Answer Competitions
|
91.98% ± 3.66% | 3/38 | $6.29 | 11731 |
|
CMIMC 2025
🔢 Final-Answer Competitions
|
90.00% ± 4.65% | 6/31 | $6.94 | 17108 |
|
HMMT Nov 2025
🔢 Final-Answer Competitions
|
89.17% ± 5.56% | 11/17 | $4.65 | 15483 |
|
IMO 2025
✍️ Proof-Based Competitions
|
38.10% ± 19.43% | 1/7 | $53.61 | 725147 |
|
Project Euler
💻 Project Euler
|
N/A | N/A | $28.93 | 39853 |
Sampling parameters
- Model
- gpt-5--high
- API
- openai
- Display Name
- GPT-5 (high)
- Release Date
- 2025-08-07
- Open Source
- No
- Creator
- OpenAI
- Max Tokens
- 128000
- Read cost ($ per 1M)
- 1.25
- Write cost ($ per 1M)
- 10
- Concurrent Requests
- 32
- Batch Processing
- No
- OpenAI Responses API
- Yes
Additional parameters
{
"reasoning": {
"summary": "auto"
}
}
Most surprising traces (Item Response Theory)
Computed once using a Rasch-style logistic fit; excludes Project Euler where traces are hidden.
Surprising failures
Click a trace button above to load it.
Surprising successes
Click a trace button above to load it.